Health Care Costs Related to Home Spirometry in the eICE Randomized Trial

Abstract

Background:

Home spirometry with regular symptom assessment is one strategy to track lung health to intervene early in episodes of pulmonary exacerbations (PE). In a multi-center randomized controlled trial home spirometry and symptom tracking demonstrated no significant differences regarding the primary clinical endpoint, FEV₁, compared to usual care, but did identify differences in healthcare utilization. We used data from the Early Intervention in Cystic Fibrosis Exacerbation (eICE) study to evaluate whether home monitoring of PE is a cost-minimizing intervention in the context of this randomized trial.

Methods:

We reviewed healthcare resource utilization of all 267 eICE participants, including outpatient visits, antibiotics and hospitalizations. Prices were identified in the IBM/Watson MarketScan® Commercial Claims and Encounters Databases and averaged over the 2014–2017 period. Using total healthcare utilization costs, we generated summary statistics by intervention and protocol arm (total cost, mean cost, standard deviation). We performed Welch Two Sample t-tests to determine if total costs and cost by type of utilization differed significantly between groups.

Results:

Outpatient visit costs were significantly higher by 13% in the Early Intervention (EI) than in the usual care (UC) arm ($3,345 vs. $2,966). We found no significant differences in outpatient antibiotic, hospitalization, or total health care costs between the arms.

Conclusions:

Within the context of the eICE trial, outpatient visits were significantly higher in those with experimental home spirometry care, but that did not translate into statistically significant differences of overall health care costs between the two arms.

INTRODUCTION:

Acute pulmonary exacerbations (PEs) are key clinical events for CF patients and pose a significant threat to their long term lung function¹. In 25% of PEs, lung function does not recover to baseline.^2–4 PEs have also been connected to decreased quality of life even after controlling for lung function, nutritional status and demographic factors.⁵ In addition, PE-associated healthcare costs are high, as patients experiencing acute PEs are likely to be hospitalized and receive intravenous (IV) antibiotics.^6,7 This is in the context of CF as a very high-cost disease in the US where medications can account for up to 85% of total costs; PEs requiring IV antibiotics are more than 10 times as costly as those requiring only oral antibiotics.^6,7 Developing interventions that prevent exacerbations or minimize their impact are vital for improving patient outcomes and decreasing the burdensome healthcare costs of CF care.^8,9

One strategy that has been considered to minimize the impact of PEs is the use of home spirometry [assessing forced expiratory volume in one second (FEV₁)] coupled with regular symptom tracking, in order to identify PEs as early as possible and intervene before they become severe.^10–14 In recent studies, this strategy had shown promise.^8,15 However, in the Early Intervention in Cystic Fibrosis Exacerbation (eICE) trial, a multi-center randomized controlled trial with 267 participants, home spirometry and symptom tracking demonstrated no significant differences in FEV₁ (the primary clinical endpoint at 12 months) compared to usual care.^9,16 The study did identify differences in healthcare utilization between the two study arms, including more acute outpatient visits but decreased IV medication administration in the home spirometry arm. As such, we conducted analyses to investigate whether home monitoring is a cost-minimizing intervention in the context of the eICE trial and what implications that may have for the U.S. healthcare system. Based on findings from the original eICE study, we hypothesized that the home monitoring intervention would save more costs than usual care.

Methods

Study Design and Population

The design is a non-concurrent prospective study, in that we retrospectively reviewed the healthcare costs of all 267 participants from the prospectively conducted eICE study, which has been described previously.⁹ Briefly, participants were individuals with CF at least 14 years old from 14 accredited CF centers across the US. Participants were randomized 1:1 to either an early intervention arm (EI) or a usual care arm (UC). The early intervention arm used home spirometers and measured respiratory symptoms with the CF Respiratory Symptom Diary (CFRSD) twice per week.^11,17,18 The study sites were automatically alerted if a participant’s FEV₁ fell below baseline by more than 10% or if the CFRSD worsened in two or more of eight respiratory symptoms. The usual care arm participants were instructed to notify their centers if they had symptoms of a PE.

Healthcare costs included all outpatient visits, hospitalizations, and outpatient antibiotics. Additional costs unique to the intervention, such as the home monitoring machine and staff outreach time, were also included in the evaluation.

We used both Intent-to-Treat (ITT) and Per-Protocol (PP) approaches to evaluate healthcare costs. In the ITT analysis, all study participants were included; EI subjects were compared to UC subjects regardless of which treatment arm they experienced. In the PP analysis, we compared the costs of those in the EI arm who adhered to the protocol and those who did not. Participants were considered adherent if they submitted one home monitoring reading per week for at least 80% of their follow-up weeks.

Determining Costs

Intervention Costs

For the EI arm participants, the one-time fixed cost of $435 for the Viasys AM2 home spirometer was included (direct purchase price to eICE study). We use the term costs for convenience: most of the estimates are closer to billing amounts or payments rather than formal evaluation of the cost of services; this is typical for health economic analyses. Variable costs associated with the intervention included the time that Respiratory Therapists would spend by telephone orienting patients to the home spirometer and providing reminders to perform home monitoring. The dates of these calls were recorded by the eICE study; however, since duration was not captured, calls were assumed to last ten minutes. Using the median salary of a Respiratory Therapist in 2018 of $60,280, we estimated that a 10-minute call cost would cost $4.83.¹⁹ We selected the role and salary of a Respiratory Therapist as they would be the appropriate staff person to perform outreach should the home spirometry intervention be implemented more widely into clinical care. The cost of each call was multiplied by the number of calls each participant received.

Healthcare Utilization Costs

Healthcare utilization costs included inpatient hospitalizations, outpatient visits, and outpatient antibiotics. Costs for these variables were identified in the IBM/Watson (formerly Truven Health Analytics) MarketScan^® Commercial Claims and Encounters Databases from 2014–2017 period, which mimicked the years of the eICE study.

Hospitalization and outpatient visits costs were identified among people with a CF diagnosis, pulmonary exacerbation diagnosis codes (ICD-9 diagnosis code 277.02 and ICD-10 diagnosis code E84.0) from 2014–2017 and outpatients were those with ICD-9 diagnosis code for CF (ICD-9 diagnosis code 277.0X; ICD-10 diagnosis code E84.X). When trying to determine the outpatient antibiotic costs, we first searched for drug costs among patients with a CF diagnosis. If patients with CF were not identified, we then searched among patients without a CF diagnosis; 32.5% (n=104) of outpatient antibiotic costs were identified among patients without a CF diagnosis.

Hospitalizations

To determine the price of a hospitalization among patients with CF, we used the total gross payments made to providers who submitted claims for covered services during an admission. Mean prices varied by length of stay (LOS) ranging from $17,857.62 (n=139, LOS: 2 days) to $1,026,343.48 (n=1, LOS: 118 days). Using the eICE dataset, we calculated the duration of each hospitalization, and used the corresponding mean Marketscan price for that LOS. We then totaled the hospitalization costs for each participant.

Outpatient Visits

To determine outpatient visit costs, we queried Marketscan for outpatient visits with a CF primary diagnosis code (ICD-9-CM: 277.0, 277.02; ICD-10: E84.0, E84.9). The mean cost per outpatient visit was $522.03 (n=190,987). Using the eICE dataset, we determined the total number of outpatient visits for each participant. This included all quarterly appointments, any outpatient PE visits, follow up appointments after PEs, and early withdrawal visits. To calculate each participant’s total outpatient visit cost, we multiplied the mean price by their total number of outpatient visits.

Outpatient Antibiotics

Although the eICE study recorded every prescription ordered during the study, we excluded medications ordered during hospitalization under the assumption that those costs were included in the hospitalization cost per the patient’s Diagnosis Related Group (DRG).²⁰ Non-antibiotic prescriptions ordered from the outpatient setting were excluded because they are frequently prescribed for non-CF conditions and are likely unrelated to the provision of CF care. Of the outpatient prescriptions ordered in the eICE study, 32% were antibiotics. We identified 320 unique antibiotic names, doses, and routes of administration.

To determine the cost of outpatient antibiotics, we identified a corresponding NDC code for each prescription dose and route of administration. We initially searched each drug’s NDC code in the Marketscan database among patients with a CF diagnosis and determined the cost of the drug. If no drugs were found among the CF population, the search was repeated among patients with any diagnosis from 2014–2017. The majority (67.5%) of all prescription costs were found among patients with CF. We then determined the price per milligram or milliliter of each NDC code, multiplied by the daily amount and then multiplied that by the prescription length in days and totaled. Chronic antibiotics and antibiotics taken as maintenance were included in our calculations. Chronic antibiotics were indicated as such in the original prescription data. We extended the prescription over either the entire study or the remainder of the study starting with the prescription date, as appropriate. If something was noted as being prescribed every other month, we took that into account. Example, one where inhaled tobramycin is prescribed ‘twice daily’ incurred twice the cost of the same drug as prescribed ‘twice daily, alternating months’.

Statistical Analyses

Descriptive Analysis

We performed descriptive analyses on the study participants, including baseline clinical characteristics, rates of healthcare utilization, and participant-reported symptoms related to lung health, as recorded by the Chronic Respiratory Infection Symptom Score (CRISS) questionnaire. The CRISS is a validated 8-question survey designed to detect symptoms of pulmonary exacerbations (Appendix A).^11,21 A higher score indicates larger disease burden.

We used descriptive statistics including means and standard deviations (SDs); non-normally distributed variables were reported as medians with interquartile ranges (IQRs). We generated summary statistics by intervention and protocol arm (total cost, mean cost, standard deviation, median cost, interquartile range) using the total healthcare utilization cost associated with each participant. Because our goal was to establish population costs, we compared mean and SD to describe costs as the primary outcomes.

Inferential Analysis

We performed Welch Two Sample t-tests to determine if total costs and cost by utilization type differed significantly (α < 0.05) between groups. We also examined median costs and utilization, where Mood’s median test was performed post hoc to test differences in total costs by intervention arm.

Typically, in a cost-effectiveness analysis the denominator would include the difference between two study arms in life-years or quality adjusted life-years. However, the eICE trial did not demonstrate significant differences that would lend themselves to such characterization. Therefore, we used differences in the CRISS scores to estimate the cost it would take to achieve the differential outcome by treatment protocol. All p value analyses were performed using a two-sided 0.05 significance level. All analyses were completed in R version 4.0.0 [R Core Team (2013)].²²

Results

Demographic data

Table 1 shows the similarity between the EI and UC arms. Both study arms have about 30% under 18 years of age, evenly divided by gender, and about 13–14% with severely impaired lung function. Although some differences appear substantial, none reach the conventional p < 0.05 threshold for statistical significance.

Table 1.

Descriptive analysis of groups.

	ITT (all participants)		EI arm:
	EI arm (n=135)	UC arm (n = 132)	Adherent to Intervention (n = 67)	Non-Adherent to Intervention (n=68)
% Pediatric (<18 years)	28.1%	29.5%	31.3%	25%
% female	50.4%	51.5%	50.7%	50%
% Pancreatic sufficient	19.3%	14.4%	25.4%	13.2%
Mean FEV1 % pred. (SD)	80.02 (22.81)	79.04 (24.6)	82.77 (24.1)	77.3 (21.11)
Disease severe <50% predicted FEV1	13.3%	14.4%	13.4%	13.2%
Disease moderate 50–75% predicted FEV1	24.4%	23.5%	22.4%	26.5%
Disease mild >75% expected FEV1	62.2%	62.1%	64.2%	60.2%
Mean change in CRISS score (SD), n	0.74 (13.01), n=129	3.55 (12.81), n=130	0.61 (13.71), n=66	0.89 (12.34), n=63

Health care utilization data by arm

Table 2 shows a slight but significant increase in outpatient visits between the EI arm (6.24) and the UC arm (5.61). Similarly, there is a significant increase in volume of PE visits for the EI arm (1.82) vs the UC arm (0.97). Hospitalizations per participant were slightly lower, though not significantly, for the EI arm (0.53) compared to the UC arm (0.63).

Table 2.

Utilization data.

	ITT (all participants)		EI arm
	EI arm (n=135)	UC arm (n = 132)	Adherent to Intervention (n = 67)	Non-Adherent to Intervention (n=68)
Outpatient visits	6.24 (3.18)*	5.61 (1.87)*	7.24 (3.12)***	5.26 (2.95)***
Routine care visits	4.18 (1.3)*	4.51 (0.96)*	4.66 (0.82)***	3.72 (1.51)***
Acute PE visits per participant	1.82 (1.96)***	0.97 (1.20)***	2.33 (2.31)*	1.32 (1.35)*
Post-PE follow up visits per participant	0.93 (1.35)*	0.61 (0.92)*	1.12 (1.59)	0.73 (1.04)
Hospitalizations per participant	0.53 (1.05)	0.63 (1.23)	0.38 (0.96)	0.68 (1.1)
Hospitalization rate (hospitalizations/person-years)	0.595	0.64	0.41	0.80
Days hospitalized per participant	3.97 (8.95)	3.95 (8.69)	3.10 (8.4)	4.82 (9.31)
Days hospitalized per participant per year of follow-up (days hospitalized//per son-years)	4.43	4.06	3.27	5.71

Mean (SD); significance

^***=p<0.001,

^*=p<0.05

Pharmaceutical use by arm

Given the variability in cost between IV, inhaled and oral antibiotics it is essential to consider if prescription patterns differed between the various groups. Although the EI group was prescribed more oral antibiotics in total (Table 3), the average days a patient was prescribed oral antibiotics was less than the UC group. The total number of IV drug prescriptions in the EI group was slightly less than UC, though with longer courses on IV drugs. While the UC group was prescribed more inhaled antibiotics, the average course duration between the groups was similar.

Table 3:

Outpatient Antibiotic Use by Arm

	EI arm (n = 135)	UC arm (n = 132)
Antibiotic prescriptions (all routes)	986	969
Duration in days (mean, SD)	60.78, 103.68	62.00, 112.45
IV antibiotic prescriptions	51	53
Duration in days (mean, SD)	19.18, 13.66	14.38, 3.85
Inhaled antibiotic prescriptions	418	470
Duration in days (mean, SD)	34.43, 40.15	31.50, 25.69
Oral antibiotic prescriptions	510	437
Duration in days (mean, SD)	27.99, 43.29	32.47, 85.18

While there appear to be slight differences (data not shown) in the total number of prescriptions by route and the average days prescribed, we did not find evidence of significantly different prescription patterns between treatment arms or between protocol adherence levels.

Cost results

Table 4 shows the mean healthcare utilization costs by treatment arm and utilization type. Total mean costs were lower in the EI arm (mean: $66,916) than in the UC arm (mean: $75,989). The difference in total costs between treatment arms is primarily accounted for by differential antibiotic costs, which were lower in the EI group. While the difference in total costs between treatment arms – a 35% increase in total costs from EI to UC – may seem substantial, the difference was not significant.

Table 4:

Mean Costs.

	ITT (all participants)		EI Arm
Mean costs	EI arm (n=135)	UC arm (n = 132)	Adherent to Intervention (n = 67)	Non-Adherent to Intervention (n=68)
Hospitalizations	$19,040.76 ($40,297.69)	$20,353.18 ($42,605.43)	$14,289.88 ($37,990.50)	$23,731.79 ($41,925.19)
Outpatient visits	$3,344.86 ($1577.38)*	$2,966.08 ($934.09)*	$3,833.41 ($1,568.82)***	$2,863.49 ($1,431.02)***
Antibiotics	$44,042.74 ($110,526.95)	$52,670.00 ($154,376.46)	$61,498.48 ($146,474.22)	$26,843.69 ($50,175.15)
Home Spirometer	$435 (0)	$0.00 (0)	$435 (0)	$435 (0)
RT outreach	$52.84 ($83.66)***	$0.00 ($0.00)***	$73.24 ($110.63)*	$32.74 ($31.98)*
Total average costs	$66,916.21 ($116,605.42)	$75,989.26 ($162,416.92)	$80,130.02 ($148,286.04)*	$53,896.71 ($70,617.04)*

Mean (SD); significance

^***=p<.001,

^*=p<.05.

Though the adherent group within the EI arm exhibited 40% lower hospitalization costs ($14,290 vs. $23,732), this was more than compensated for by much higher antibiotic costs ($61,498 vs. $26,844.) This resulted in overall higher average costs of $80,130 for adherent participants compared to $53,897 for non-adherent participants.

Given healthcare costs often have a skewed distribution, we reanalyzed the cost data using median costs. Despite such skewed distributions, we found the same pattern of costs by intervention arm (Table 5). Differences in median total costs were not significant.

Table 5:

Total Median Costs. Median (IQR).

	ITT (all participants)		EI Arm
Median costs	EI arm (n=135)	UC arm (n = 132)	Adherent to Intervention (n = 67)	Non-Adherent to Intervention (n=68)
Hospitalizations (IQR)	$0.00 (20,178.21)	$0.00 (20,279.45)	$0.00 (0.00)	$0.00 (26,650.3)
Outpatient visits (IQR)	$3,132.18 (2,088.12)	$2,610.15 (1,044.06)	$3,132.18 (2,349.13)	$2,871.17 (2,088.12)
Antibiotics (IQR)	$4,586.83 (40,278.76)	$2,785.36 (42,988.97)	$4,605.77 (47,725.37)	$3,764.06 (33,521.48)
Home Spirometer (IQR)	$435 (0.00)	$0.00 (0.00)	$435 (0.00)	$435 (0.00)
RT outreach (IQR)	$33.81 (50.72)	$0.00 (0.00)	$43.47 (50.72)	$19.32 (43.47)
Total median costs (IQR)	$25,574.06 (74,199.49)	$23,891.48 (75,439.88)	$25,487.24 (80,803.58)	$26,945.72 (67,233.30)

Changes in CRISS score and related costs

While there were no statistically significant differences in FEV₁, there was a significant change in the CRISS score, with a change of 3.55 points among the UC arm and 0.74 among the Intervention arm participants. Within the eICE study, achieving that change in CRISS cost $9,073.05.

DISCUSSION

For over a decade, analysts have described an increasing use of medical technology in the home environment.^23,24 A major driver of this increase has been the high cost of health care, particularly inpatient hospitalization. However, other factors have contributed, including the recognition by the medical technology industry that patients and caregivers can effectively use sophisticated technology, and that remote monitoring facilitates greater patient safety. More recently, the COVID-19 pandemic further heightened the need to keep patients at home. The high rate of hospitalization for pulmonary exacerbations among CF patients and the presence of home spirometry as a technology presented clinicians and researchers with an excellent opportunity to determine the effects of utilizing this home-based technology to improve CF care.

As noted earlier, early feasibility studies had suggested that home spirometry had the potential to reduce hospitalizations for these patients. In fact, a study that was published after the results of the eICE study showed the potential to improve medication adherence in this population and continued to suggest a potential for home monitoring to have a positive clinical impact.^8,15

While the original eICE trial was negative for its primary outcome, which should temper enthusiasm for adopting home spirometry, the COVID-19 pandemic has caused a dramatic increase in telehealth. Home spirometry is now being utilized widely in CF and this is likely to continue after the pandemic is over. Therefore, the additional information about healthcare utilization in the current study is both timely and relevant for the CF community. In addition, although the study arms had similar lung function after a year with narrow confidence intervals, care did differ. Understanding the implications of such a change in care could influence care particularly if home monitoring had been cheaper and if it demonstrated a value proposition for home monitoring.

We anticipated higher rates of outpatient visits for the EI arm participants as the home spirometry measures FEV₁ and subsequently prompts patients to seek care based on its findings. Likewise, we expected the intervention to decrease the rate of hospitalizations, as the act of early intervention should reduce the severity of PEs and the need for inpatient hospitalization. This did not turn out as we had expected largely because of the large variability in health care costs in our sample. For example, we expected the intervention group to have higher outpatient costs, which occurred ($3,344 vs. $2,966), and this did result in lower than expected hospitalization costs ($19,041 vs. $20,353), and in lower antibiotic costs ($44,043 vs. $52,670). Only the first of these three showed a statistically significant difference and, thus, the overall cost difference was not significant between the EI and UC arms. High levels of variability are common in health care cost studies; thus, we attempted to further analyze these differences by examining medians, which also did not reveal statistically significant differences.

We further attempted to unravel cost and utilization patterns within the intervention group because of known incomplete adherence with the study protocol. We compared the adherent to non-adherent group and discovered rather substantial differences. The adherent group had significantly higher outpatient costs, but no differences in hospitalizations or antibiotic costs. Once again, variability prevents us from declaring differences, such as the more than double antibiotic costs of the adherent group compared to the non-adherent group, as significantly different.

We did look at the improvement based on the CRISS and showed a change in one unit of improvement comes at a cost of $3,200. This is worth noting as a possible benefit of this approach; however, the CRISS does not map into health status preventing the calculation of the usual quality adjusted life year calculation standard in cost-effectiveness analysis.¹⁰

One limitation we note is the relatively modest sample size that might lead us to conclude a lack of differences in patterns of cost and utilization. The very large variability in cost in this sample may have led to a type 2 error. Further, some costs are missing, for example, we were not able to code non-antibiotic drug costs, because the trial limited collection to those medications more directly related to CF lung disease and the treatment of exacerbation. Any encounter with CF providers could lead to changes in CF medications; we did not specifically look at changes in CF medications because of the intervention. We do not have evidence this was a common occurrence, but we recognize this as a limitation. Though these are likely not largely different between the arms, they nevertheless could change the picture of the study sample. Furthermore, by examining costs from a societal perspective, we used costs attributed to hospital stays measured by a standard Diagnosis Related Group (DRG) charge, which meant that if the hospital stays for CF patients include a substantial amount of expensive IV antibiotics, we were not able to separate those costs and thus missed potential differences between the arms. A cost analysis using the perspective of the hospital that accounts for costs of expensive inpatient antibiotics might reveal different patterns or tradeoffs that home spirometry would support.

CONCLUSIONS

Despite expected shifts in outpatient care that seemed to reduce inpatient stays for those using home spirometry monitoring, especially among those CF patients who adhered to the study protocol, there were no statistically significant reductions in overall health care costs for the home spirometry intervention group in a randomized controlled trial. Not only does home spirometry not improve lung function or reduce subsequent pulmonary exacerbations, cost and health care use do not appear to provide a justification for this practice. It is conceivable that the patterns that we observed may have different implications in healthcare systems that include all costs for patients, including all inpatient care where heavy use of IV antibiotics would cost the health care system. This analysis could lead to targeting high value care and de-emphasizing lower value care for the CF community of patients, families, and clinicians.

Highlights.

• Home spirometry can help track lung health to intervene early in pulmonary exacerbations

• The eICE trial showed no differences in FEV₁ but identified health care use differences

• Outpatient visit costs were significantly higher comparing users to non-users of home spirometry

• However, overall health care costs were similar in users and non-users of home spirometry

Abbreviations:

CF

Cystic Fibrosis

FEV₁

Forced expiratory volume in one second

PE

Pulmonary exacerbations

eICE

Early Intervention in Cystic Fibrosis Exacerbation

EI

Early Intervention

UC

Usual Care

CFRSD

CF Respiratory Symptom Diary

CRISS

Chronic Respiratory Infection Symptom Score

DRG

Diagnosis Related Group

NDC

National Drug Code

LOS

Length of stay

Appendix A: CF-RSD: The CRISS questions include #1- #8 of the CF-RSD: